Pronoun Morphology

― ―1

Stephen Howe

Introduction representing properties by morphological patterning – i.e. shared properties are indicated by shared formatives This paper examines the degree of regularity-irregularity – or, on the other hand, marking differences in property and the complex morphological type of the personal pro- by suppletion – where a personal pronoun is morphologi- nouns in English, but also examines the other Germanic cally distinct from other pronouns with which it shares a languages and Japanese. property or properties. The paper draws largely on my PhD thesis (Howe The paper also discusses how accented and unaccented 1996) on personal pronouns in the Germanic languages, forms of the same pronoun can vary in their connection summarizing the main theoretical points on morphol- to one another. Not only can the personal pronouns show ogy and change. Parts of this paper will be presented at a suppletive or suppletive-like distinctions between sepa- conference on Irregularity in Morphology (and Beyond) rate pronouns, i.e. not derivable by general synchronic in Bremen, Germany, in autumn this year. rule, but also non-synchronically-derivable variants of The morphology of the pronouns with considerable the same pronoun may occur. irregularity is not well accounted for by morphological The morphology of the personal pronouns is in many theories that concentrate on regularity. This paper, as well cases grammatically, semantically and formally complex. as examining personal pronoun morphology in English The personal pronouns in English and other Germanic and other languages, attempts to account for this complex languages are primarily representative, portmanteau morphology. forms rather than active indicators of each category/ A fundamental characteristic of the personal pronouns property: one personal pronoun cannot usually be de- in the Germanic languages – and an important one for rived from another just as one lexeme cannot usually be personal pronoun morphology – is that they are short. A predicted from another. Personal pronouns are generally main function of personal pronouns and other proforms (co)referring terms, both grammatically and semantically is to abbreviate. A further characteristic is that the per- to the external world – (in their core meaning) ‘I’ = the sonal pronouns are generally among the most frequent speaker, ‘we’ = a group to which ‘I’ belong, ‘he’ = the words in English and other Germanic languages – virtu- male person etc. – and therefore it is perhaps not surpris- ally all the subjective and objective pronouns in English ing that also formally the personal pronouns show simi- for example occur in the first one hundred most frequent larities both with regular morphology and with lexical or words in speech. One of the most obvious consequences content words. This grammatical-lexical duality will be of high frequency is that a frequent form is more likely discussed further in the paper. to be short rather than long. However, it is important to distinguish between shortness and ambiguity – the rele- The personal pronouns in connected speech vant factor is not how short a personal pronoun is, rather whether or not it is ambiguous. Ambiguity as a factor in For pronouns and other proforms, as stated above, a fun- change in the personal pronouns will be taken up later in damental function is to abbreviate. It is the raison d’être the paper. of the personal pronouns to be (relatively) short – there Theoretically, the morphology of the personal pro- would be little point in personal pronouns being as a rule nouns is analysed in this paper as representing two dif- longer than the noun phrases they substitute (including ferent systematic types: either systematic in terms of for the 1st and 2nd person terms such as name and title).1 marking property connections, or systematic in terms The English personal pronouns are function words. of marking property differences. Either on the one hand This means that, like the, a, be, that, on and Japanese wa,

（ 1 ） ― ―2 福岡大学研究部論集 A ９（７） 2009

ga, no and o for example, they have grammatical func- clear, intelligible, continuous speech: when these were tion. Although usually written in their orthotone forms, played to listeners, there was great difficulty in making a the personal pronouns in English are mostly unaccented correct identification. Crystal states that ‘Normal speech in normal connected speech. A study of the personal proves to be so rapidly and informally articulated that in pronouns in English must therefore take account of their fact over half the words cannot be recognized in isola- variation in accent2 and consider them in connected tion.’ Gimson/Ramsaran (1989: 290) state on function speech (cf. Howe 1996). The isolated written or citation words that ‘Such is the reduction and obscuration of the form of a personal pronoun is in connected speech the unaccented forms that words which are phonetically and exception rather than the rule. For example, Gimson/ phonemically separate when said in isolation may be Ramsaran (1989: 26) state that his, her, we and them neutralized under weak accent.’ They add that ‘Such neu- have over 90% occurrences as unaccented forms. It is tralization causes no confusion because of the high rate important to bear in mind, then, that for English the writ- of redundancy of meaningful cues in English; it is only ten language generally does not represent the most usual rarely that the context will allow a variety of interpreta- forms of the personal pronouns in speech.3,4 tion for any one cue supplied by an unaccented word Function words differ considerably from lexical words form.’ in connected speech – compare again Gimson/Ramsaran The important difference in accent between personal (1989: 265f.) for English: ‘Content words … gener- pronouns (and similar function words) in English and ally have in connected speech the qualitative pattern of lexical or content words is immediately apparent in a their isolate form and therefore retain some measure of comparison of the personal pronouns with (partially) ho- qualitative prominence even when no pitch prominence mophonic lexical words: is associated with them and when they are relatively unstressed.’ Many function (or ‘grammatical’ or ‘form’) eye mine words, on the other hand, have ‘two or more qualitative yew ewe yaws and quantitative patterns according to whether they are wee hours unaccented (as is usual) or accented …’. As Gimson/ Ramsaran (1989: 261) point out, function words in Eng- hymn lish such as the personal pronouns, articles and auxiliary verbs are likely to be unaccented, although they may be The difference is similarly apparent when personal pro- accented if the meaning requires it. On connected speech nouns are used as nouns in examples such as ‘Is it a he and function words in Japanese, see Shibatani (1990: or a she?’, ‘You’re it’ (in children’s games), ‘The diet to 175–177) and Tsujimura (2007: 92–94). create a new you’ etc.5 Further, there may be evidence In a study of connected speech (cited in Crystal, 2003: for a psychological and neurological distinction between 147), single words were cut out of a tape recording of function and content words (e.g. Fromkin & Rodman

１ Formality and its influence on length will be discussed later in the paper. ２ In this paper, ‘accent’ is used as in Gimson/Ramsaran (1989) where ‘variations of pitch, length, stress, and quality, contribute to the manifestation of the accented parts of connected speech’ (1989: 262). Generally accent variants in the personal pronouns will be referred to as accented and unaccented where this is unambiguous, and by a convention +accent(ed) and −accent(ed), which represents greater–less accent(ed) (and not necessarily straightforwardly with/without or plus/minus accent). The use of the variables + and − accent(ed) – i.e. relative rather than absolute terms – is very useful in cross-linguistic study where absolute dichotomous terms are sometimes less helpful. Note that the use of accented–unaccented or + and − accent(ed) should not be taken to mean that there are necessarily only two accent variants. ３ Written English has only the contracted form ’s (Let’s go) and the archaic ’t (’Twas). As well as unaccented pronouns, some languages have specific emphatic forms, such as ikke (‘I’) in Dutch, Frisian and German dialect. ４ Genitive/possessive forms, not the main focus here, are always accented in English (see Quirk et al. 1985: 362). ５ In Japanese, the words for ‘he’ and ‘she’ – kare and kanozyo – are used for boyfriend and girlfriend respectively. Further examples of lexicalized pronouns in English are thou (thee, thine, thy) and ye: for the majority of Present English speakers, these forms are not part of their usual pronoun system, though they are still known, and may be used, as pronouns. Significantly, however, they may lose the accent variation typical of personal pronouns and other function words, occurring only in their citation form. Forms such as thou etc. and ye can be said to have been lexicalized – i.e. although they retain the pronoun form, they more resemble lexical words than function words. Note further that in the lexicalization of pronoun forms discussed here, the loss of accent variation characteristic of English personal pronouns and occurrence only of the citation form mirrors one of the processes cited by Hopper & Traugott (1993: 2f.) as typical of grammaticalization, namely phonological reduction (of auxiliaries) as in for example going to > gonna, or will > ’ll.

（ 2 ） Pronoun Morphology（Howe） ― ―3

1993: 39, 440 & 445). However, as touched on in the in the London–Lund Corpus of Spoken English (= edu- introduction, the pronouns have an important duality cated British English),6 all the subjective and objective of grammatical and real-world categories and reference pronouns except us occur in the first one hundred most which will be discussed further below. frequent words: I is 3rd, you is 7th, it is 10th, he is 18th, In Japanese, we will maintain, the equivalent of Eng- we is 23rd, they is 24th, she is 59th, me is 66th, them is lish unaccented pronouns is zero – i.e. where reference 77th, him is 89th and her is 96th. Us occurs, according is clear, English speakers use an unaccented form of the to Gimson/Ramsaran (1989: 266), in the first 200 most pronoun and Japanese makes no overt (co)reference. common words in connected speech. The reference of unaccented pronouns in English – i.e. As pointed out by Zipf (cited in Mańczak, 1980: 50), to given, anaphoric or indefinite referents or antecedents one of the most obvious consequences of high frequency rather than new, focus, or contrastive reference – can per- is that a frequent form is more likely to be short than haps be termed agreement and in other languages may be long. In language, shortening of frequently used words absent – as indeed we find in Japanese (cf. Howe 1996: is common, for instance in English PC, phone, flu, TV 55, Tsujimura 2007: 254–257). or telly, and in Japanese pasokon, rimokon, makku and Thus, we can see in the ‘over 90% occurrences’ of sûpâ. many English personal pronouns as unaccented and in Although personal pronouns have very high frequency the common ‘zero’ pronouns of Japanese a parallel or in English, in Japanese, by contrast, Suzuki (1978: 113) equivalence. states that ‘investigations into actual usage make it clear Given that where reference is clear from context (text that personal pronouns appear only on very limited oc- or situation) a pronoun will normally be unaccented in casions’. One difference between English personal pro- English and in Japanese zero, it is not surprising that nouns and some Japanese pronouns is length. As a clear Japanese pronouns, when they do occur, occur mostly in example, contrast English I with its very formal Japanese orthotone form. According to Hinds (1986: 248), in Japa- equivalent watakusi. Here we must add that, although the nese ‘There is no difference in segmental or supraseg- length of a pronoun is basically a function of its frequen- mental structure of pronouns depending on whether the cy, formality is also a factor, especially so in Japanese, context is emphatic or unemphatic. Nor is there a differ- with more formal forms, if used, tending to be longer ence in accentuation, tone variation, or vowel length.’ – compare the Japanese forms for ‘I’ ore, boku, atasi, However, contraction of Japanese pronouns does occur watasi, watakusi.7 Conversely, the more informal, and and will be discussed later as well as in a forthcoming likely also the more frequent, the shorter a form tends to paper (Howe forthcoming a). be. Two forthcoming papers, on ‘Reference and ellipsis’ and ‘Pronouns and politeness’ (Howe, forthcoming c and Morphology and frequency d), will examine why Japanese pronouns are less frequent than English pronouns. As stated in the introduction, in English the personal pronouns are among the most frequent words. For example

６ Svartvik et al. (1982: 43–46) ７ This is also often the case in language generally, where formal or polite language tends to be less direct, more elaborated and longer than informal language. Compare the following examples in Japanese (from Bunt 2003: 213–223 or my own) showing differences in word choice, morphology, titles and utterance length: da – desu – de gozaimasu iku (-anai, -ta) – ikimasu (-masen, -masita) etc. o-kyaku-sama Genki? – O-genki desu ka? Jun-chan, mô tabeta? – Sensei wa mô mesiagarimasita ka? And English: Shut up! – Please be quiet – Would you mind not talking Got the time? – Excuse me, could you tell me the time, please? One could state, then, that frequency and formality are opposite tendencies, one to abbreviate, the other to elaborate. If we suggest that unaffected, informal language tends to shortness and economy, that utterances are lengthened when formal would seem to confirm this assumption – speakers go to some length to be polite. A third relevant factor is pragmatic constraint, for example where titles and/or names are used in place of, or in avoidance of pronominal address. This is common in Japanese and is also found in English and other European languages. Pragmatic constraint, including taboo, will be discussed briefly later and further in Howe (forthcoming d).

（ 3 ） ― ―4 福岡大学研究部論集 A ９（７） 2009

Patterning and suppletion rather than formally indicating each property individually. The forms in Figure 3 are arbitrary representative Figures 2 and 3 illustrate schematically two possible terms in the same way that most lexemes have an ar- morphological types: the first is a perfectly patterning ag- bitrary relationship between form and meaning. These glutinating paradigm, and the second a perfectly supple- pronouns have no formal connection to one another and tive portmanteau paradigm.8 a change in a property will result in a complete change of form. The forms in Figure 3 cannot then be generated by Figure 2: Patterning paradigm rule. 1SN 1SO 1SG One consequence of the high frequency discussed above is that frequent forms are more likely to be sim- 2SN 2SO 2SG plex, ready-made forms, as in Figure 3, rather than com- 3SMN 3SMO 3SMG bined from separate elements each time they are required. 3SFN 3SFO 3SFG A portmanteau form, as well as being ready-made, may 3SXN 3SXO 3SXG possibly also be shorter than an agglutinating form made 1PN 2PO 2PG up of several different elements: where a portmanteau form requires one morph only, an agglutinating form, as 3PN 3PO 3PG in Figure 2, requires several elements.

Figure 3: Suppletive paradigm Patterning X H E If we examine the actual personal pronouns in English, S D K however, they are not as completely isolated as the forms Y U O in the hypothetical suppletive portmanteau paradigm. G N R There appears to be some kind of patterning in some of C L W the pronouns which, although it does not reach the level B P V of predictable full-scale inflection, may still show some potentially significant correspondences. M F A A clear indicator of the significance of patterning is analogical extension: if a formative is extended from one In the patterning agglutinating paradigm in Figure 2, or more forms to another form or forms sharing the same each form consists of a number of regular discrete ag- property or properties, we can say that in the mind of the glutinating morphs. Each morph is the same throughout language user(s) there is a connection. An example of the paradigm and unambiguously indicates its particu- such a development in English is the preliterary extension lar property, i.e. there are no allomorphs and there is a of initial h- in the (orthotone forms of the) 3rd person perfect one-to-one relationship of form to meaning. The pronouns, remnants of which can still be seen in Modern meaning of each pronoun is a function of the meaning English he, him, his and her. The subsequent – suppletive of its component parts. The ordering of the elements is or suppletive-like – developments of she and they, which entirely predictable and each element is clearly segment- will be discussed later, have obscured this pattern. In Old able. A change of a property (e.g. from 1st to 2nd person English, all the 3rd person personal pronouns, singular or from singular to plural) means a change of one morph and plural, were marked with initial h- (see Howe 1996: only, not a change of the whole form. The personal pro- 83–85 and 131–133), as comparison with two closely re- nouns in Figure 2 can thus be generated by rule. lated languages, Old Saxon and Middle Dutch, shows: In contrast, in the suppletive portmanteau paradigm in Figure 3, each personal pronoun is a single unique portmanteau morph which refers to the bundle of properties

８ For convenience, the categories/properties used here for illustration are based loosely on Modern English. In this study ‘category’ and ‘property’ are used as in Matthews (see 1974: 66 & 136) where ‘categories’ are e.g. person, number, case etc., and ‘properties’ are individual terms of categories, such as 1st, 2nd, singular, plural, nominative, accusative etc. For a survey of other terms in use see Carstairs-McCarthy (1992: 196f.).

（ 4 ） Pronoun Morphology（Howe） ― ―5

Old Saxon 3rd person pronouns elements formatives. By formatives Pike means elements h e ina im is which do have some signalling function, but which cannot always be dealt with in a conventional morphemic it it im is approach, as Pike states (1965: 219): ‘obvious formative siu sia iru iro groups are present, functioning as formal signals, but … sia sia im iro classical morphemics cannot segment these neatly because of limiting assumptions … concerning the relation Middle Dutch 3rd person pronouns of form to meaning’. Pike takes the term formative from h i h em h em sijns Bolinger (1948), but uses it differently; for Bolinger a formative is a type of morpheme that can enter into new h et h et h em (sijns) combinations, as opposed to one that has only diachronic si h aer, -se h aer h aer value. Other possible terms are submorpheme (see Crys- si h em, -se h em h aer tal 1991: 224), which is perhaps too fixed to allow for a range of relevance, and semimorpheme – Quirk et al. Old English 3rd person pronouns (1985: 1584) speak of the ‘semi-morphological status’ h e h ine h im h is of e.g. /sn/ in sneer, snide, snoop, or the ending of rattle, sizzle, tinkle. Haas (1966: 129), in a section entitled ‘Rel- h it h it h im h is evance without Contrast’, states that ‘there are important h īo h īe h ire h ire grammatical elements which contract no contrasts or do h īe h īe h im h ira so only rarely’ and ‘when, on the grammatical level, we have obtained all the distinctive elements …, we are left Japanese, too, shows similar patterning, most obvi- with a residue of important non-distinctive, or practically ously in the ko–so–a–do deictics and associated inter- non-distinctive elements’. rogatives: Hockett (1987: 97) states that there is no neat bound- ary separating strong associations from those features or Japanese ko–so–a–do patterns that give rise to vaguer associations; he believes kore sore are dore (1987: 88) that there are no objective grounds for distinction between ‘official grammatical structure’ and ‘acci- koko soko asoko doko dental’ similarity, rather ‘it is a difference of degree, not kotti sotti atti dotti of kind’. Hockett terms this kind of similarity ‘accidental’, kotira sotira atira dotira but as shown by English h- above and other examples in kono sono ano dono Howe (1996), in the personal pronouns at least such sim- konna sonna anna donna ilarities in form are in many cases not mere coincidence. kô sô â dô In this paper, as in Howe (1996), we will use the term morphological patterning to define where there is some form-to-meaning correspondence. This term allows us Other examples of patterning in English are the, this, to speak of (grammatical or semantic) form-to-meaning that, these, those, there, then, and the interrogatives correspondences that are not necessarily ‘regular’ in the what, which, where, when, why, whether and (in writing) conventional linguistic sense, but are nevertheless pres- who, whose, whom. Earlier English and other Germanic ent. Language can show significant patterning without languages have a deictic and interrogative pattern here– being derivable by rule. there–where, hence–thence–whence, hither–thither– whither paralleling the Japanese equivalents above. And Suppletion such patterning is by no means unknown in other areas of the language, for example (as also again in many re- Conversely, an element may be perceived to have mean- lated languages) the n- of the negatives no, not, n’t, none, ing not because it shows a pattern, but because it is never, neither, nor, nobody, no-one, nothing, nowhere, unique in a paradigm. Patterning and uniqueness can be non-, nil, null, nought and negative. said to represent two different systematic morphological Pike, in a discussion of German, terms such patterning types: either systematic in terms of indicating property

（ 5 ） ― ―6 福岡大学研究部論集 A ９（７） 2009

connections, or systematic in terms of marking property hour hours differences. Obviously this second type cannot denote hymn hymns ‘derived by rule’, but it does represent a systematic type. These two morphological types were illustrated in the yew yews figures above: Figure 2 represents all its properties by ewe ewes patterning, i.e. shared properties are indicated by shared formatives; Figure 3, on the other hand, rather than indi- Irregular morphology cating a connection between personal pronouns in terms Singular Plural of properties shared, marks the differences in property by I we suppletion – i.e. a personal pronoun is distinct from other pronouns with which it shares a property. mine ours Suppletion is often defined as the use of a morphologi- him them cally unrelated (though semantically related) form to you complete a paradigm, as in English go–went. An example you you + all > y’all of such suppletion in the personal pronouns in English is but indeed you you + [z] > yous(e) the 3rd person plural they (them–their), borrowed from Scandinavian. This is essentially an etymological definition of suppletion, then, but there are also other develop- And between Japanese and English plurals: ments that can result in suppletive-like distinctions, as discussed in Werner (1991) and Howe (1996). Perhaps, Japanese English then, at least synchronically speaking, it is possible to watasi-tati, -ra we extend the definition of suppletion to include all forms of a paradigm that show no connection in form even though anata-tati, -ra you (y’all, yous(e)) kare-tati, -ra they have a connection in property. In fact, in some cases they kanozyo-tati, -ra the ultimate origins of the suppletion may be uncertain; for example, although synchronically the 1st p. sing. pronouns I–me–mine, my show a suppletive subjective– Similarly, the genitive/possessives, which are formed in oblique case distinction, it is not certain whether these Japanese by addition of the particle no: forms derive from two different roots or from the same stem accented differently (Forchheimer 1953). Japanese English watasi no my, mine Complex morphology anata no your, yours The morphology of the personal pronouns is in many kare no his cases grammatically, semantically and formally complex. kanozyo no her, hers Suzuki (1978: 115) states that the ‘so-called Japanese personal pronouns’ do not form an independent word group morphologically; however, even a casual glance at The last two tables show that in the formation of the plu- the English personal pronouns shows that they are by no ral and genitive/possessive at least, Japanese pronouns means morphologically ‘regular’ either. are in fact more regular than their English counterparts. The contrast between regular inflection and the irregu- However, especially in the older stages of the German- lar morphology of the pronouns in English is illustrated ic languages, including English, the 3rd person personal below: pronouns in particular do show inflectional similarity with other pronouns, such as the demonstratives, and Regular inflection with noun phrase inflection. The connection between Singular Plural category/property distinction in noun phrases and in personal pronouns will be discussed further below. eye eyes gold etc. mine gold etc. mines

（ 6 ） Pronoun Morphology（Howe） ― ―7

Phonological developments can differ between ac- forms are not derivable synchronically. Contrast for ex- cented and unaccented pronouns ample English him, her and it below:

Accent variation in the personal pronouns can result in You’re not going to MARRY him? differences in phonological development, both between You’re not going to marry HIM? pronouns and between pronoun and non-pronoun forms. For example, a +accented form may undergo a develop- You’re not going to MARRY her? ment which the −accented form or forms do(es) not, or You’re not going to marry HER? vice versa. Important here also is possible change in the But: relative domain of originally + and − accent forms, i.e. You’re not going to EAT it? the generalization or increase in domain of one form and ?You’re not going to eat IT?9 decrease in domain of another. For example, a +accent pronoun may be generalized also as a −accented form, or There are also examples where the −accented pronoun an originally −accented pronoun may be generalized as is not synchronically phonologically derivable from the a +accent form. An example of such changes in English +accented orthotone form – i.e. the −accented form is not is the 1st person singular in Middle English ic, ik, i, ich, simply phonologically a reduced form of the orthotone with generalization if i, later ī, and Present Standard Eng- pronoun. The distinction between +accented and −ac- lish I [aɪ]. cented form can range on a scale from suppletion, such Although regular phonological developments also take as particularly well illustrated in Frisian, a language place in the personal pronouns, accent variation, sandhi closely related to English: (see Howe, 1996: 88–91) and generalization of originally +accented or −accented forms can result in different de- Frisian (Saterlandic) +accented −accented velopments to non-pronoun forms. As illustration, com- 3rd p. sing. masc. subj. hie er pare the following examples from English, again using 3rd p. sing. fem. subj. ju ze the 1st person singular ‘I’: 3rd p. plural subj. jo ze

1st p. sing. ‘tar’ To less suppletive, but nevertheless non-synchronically- Old English iċ piċ derivable differences, for example West Frisian (see Vis- Modern English I [aɪ] pitch [pɪtʃ] ser 1988: 178f. & 187f.):

Morphological, grammatical and semantic differences Frisian (West) +accented −accented between accented and unaccented forms 1st p. sing. obj. [mɛi] [mi] 2nd p. sing. T obj. [dɛi] [di] As well as separate pronouns, + and − accent forms of what we may term the same pronoun (e.g. 3rd p. plural 1st p. plural subj. [vɛi] [vi] masc. subj. ‘he’) can also vary in their connection to one another. Nübling (1992: 6f.), following Zwicky (see Visser states that there are in these forms synchronically Nübling for references), defines as a ‘simple clitic’ a neither any general phonological processes that derive clitic which corresponds synchronically to an indepen- the −accented form from the +accented form, nor con- dent full form. Such a correspondence of simple clisis versely are there any general phonological processes that accounts for many of the personal pronouns, and here we derive the +accented form from the −accented form when can speak of full and reduced forms – for example him accented. [hɪm – ɪm] and her [hɜː – hə – ɜː – ə]. However, some Further, a common development in the personal pro- personal pronouns can be described by what Nübling, nouns in the Germanic languages is reinterpretation of following Zwicky, terms a ‘special clitic’ – either the originally case forms as accent forms (see Howe 1996: clitic has no corresponding full form, or the full and clitic §2.7, §3.1.1, §3.1.3 and index for detailed references).

９ It can only rarely receive stress, for example ‘Is that IT?’, see Quirk et al. (1985: 348).

（ 7 ） ― ―8 福岡大学研究部論集 A ９（７） 2009

In such developments, the original case function of the watasi formal pronouns is lost or obsolescent, and the pronominal case atasi informal, female forms are reinterpreted as + and −accent forms. This can result in increased irregularity as such accent variants are wasi informal, older male then not simply phonologically reduced or emphatic ver- assi very informal, adult male, Tokyo Bay sions of one another. atai very informal/vulgar, female

Grammatical and semantic differences Like the case forms above, such variants show reinterpretation, though here semantic reinterpretation. Re- The fact that more information is given from context (text interpretation is common in the personal pronouns in or situation) when −accent pronouns are used means that English and other Germanic languages, see Howe (1996: in some cases unaccented forms may maintain fewer 95–100), and will be examined further, together with distinctions or have less specific reference than accented Japanese, in a subsequent paper (Howe, forthcoming b). pronouns. In Dutch, another language closely related to The discussion above shows that the correspondence English, gender, case and number reference can all vary between full and reduced forms of pronouns is not neces- between some + and − accent pronoun forms (see Howe sarily a simple one. 1996: 30–31). In English, −accented we or you can refer either defi- Relationship of categories in language with nitely or indefinitely, while accented WE and YOU refer categories in pronouns specifically and cannot have general, indefinite reference: Grammatical and real-world categories We shouldn’t watch so much television. WE shouldn’t watch so much television. A central factor in the personal pronouns is the connection between category/property distinctions in the Your country needs you. language outside the personal pronouns and those in Your country needs YOU. the personal pronouns. These categories/properties can be grammatical ones and/or natural ones based on real- An explanation for such differences is that strength of word entities.11 Examples of grammatical categories in reference is a function of accent – i.e. the more strongly the personal pronouns in the Germanic languages are accented, the stronger and thus more specific the refer- (nominative, accusative, dative, genitive) case or (mascu- ence. line, feminine, neuter) grammatical gender. Examples of In Japanese, too, contracted forms of some of the real-world-based categories in personal pronouns are for pronouns are not necessarily simply shortened variants, instance person, natural gender and T/V.12 but may have differences in register or meaning, such as Of course, categorization of the real world in language, formal or informal (with contracted forms being less for- as well as types of grammatical category, can differ from mal) or male or female, as in the feminine-labelled atasi. language to language – something abundantly clear in Makino and Tsutsui (1986/1989: 28–29), for example, worldwide comparative studies of pronoun systems – cite ‘at least’ six contracted forms of the 1st person sin- see for example the articles on pronouns by Ingram and gular, with decreasing formality:10 Head in Greenberg (1978). Further, these two types of category are not necessarily mutually exclusive – both watakusi very formal can be relevant in personal pronouns – for example in the Germanic languages the selection of 3rd person gender atakusi formal, female pronoun is frequently governed to varying degrees by

10 Further Japanese pronoun variants will be discussed in Howe (forthcoming a). 11 The distinction ‘grammatical’ versus ‘natural’ categories here is meant in the same sense as grammatical and natural gender. Both types of category are grammatical in the sense that they display formal contrasts in the personal pronouns, although governed by different criteria. 12 Socially-differentiated forms of address will be referred to in this paper as T and V (from Latin tu and vos), where T is less formal and V more formal. As will be discussed in Howe (forthcoming d), however, this terminology is not well suited to Japanese, nor is it ideal for English, as ‘T/V’ factors are also relevant for 1st and 3rd person reference.

（ 8 ） Pronoun Morphology（Howe） ― ―9 both grammatical and natural gender.13 based distinction alone, and the loss in noun phrases means that the personal pronouns are left with a gram- Grammatical categories and noun phrases matical category/property-based distinction that has little or no noun phrase parallel. Also fundamental in the personal pronouns is the con- Change is frequently a gradual process, not only in the nection between category/property distinction in noun spread in the language community, but also in the lan- phrases and in personal pronouns. The relevance of noun guage itself. A distinction may be lost in noun phrases, phrase distinction is that syntactically personal pronouns and (then) in some pronouns, and (possibly) eventually function like noun phrases. That the pronouns parallel in all forms. Similarly for example, changes in morpho- or follow distinction made in noun phrases is clear form logical to syntactic distinction take place over a long their proform nature. timescale and do not represent an either–or, but rather an This connection between category/property distinc- increase–decrease where both may be relevant. This dif- tion in noun phrases and in personal pronouns can be fusion of grammatical change is an important feature of expressed as the following implicational statement: change affecting the personal pronouns, and is one of the reasons for synchronic irregularity: change is not neces- If a category/property distinction – grammatical sarily synchronized in all forms. and/or real world – is made in noun phrases, then That person, T/V etc. can and do exist or remain as the distinction will usually also be made (though categories in pronouns even when not distinguished in not necessarily with the same formatives) in the noun phrases or even elsewhere in the language can, as personal pronouns. stated above, be explained by their real-world nature – they are not dependent on distinction made in noun Note that this implicational statement does not exclude phrases. additional real-world-based distinctions absent in noun Retention of forms does not always mean retention of phrases being made in the personal pronouns. Grammati- the original category/property, however. A common de- cal categories in the personal pronouns are dependent on velopment in the personal pronouns, touched on earlier, distinctions made in noun phrases. Real-world-based cat- is reinterpretation, where pronoun forms are reinter- egories, on the other hand, do not depend on distinctions preted into a new use. One example here as illustration made in noun phrases and can always occur; indeed, ac- already mentioned above are the English 3rd person sin- cording to Greenberg (1966: 113) person and number are gular gender forms: English no longer has a grammatical universal categories in pronoun systems (though see also masc.-fem.-neuter distinction – the personal pronouns Mühlhäusler & Harré 1990: 62–65). he–she–it are reflexes of this, but their use is governed by The distinction of both types of category outside noun different (natural rather than grammatical gender) crite- phrases, including outside the personal pronouns – for ria. example by verb morphology, syntactically, or in Japa- However, examples remain which do genuinely rep- nese by context – can be important to distinction in the resent a longer maintenance of a category/property in personal pronouns. the personal pronouns than in other word classes. One reason for the longer retention in the personal pronouns Personal pronouns often retain distinctions longer is that morphologically the personal pronouns are, as than noun phrases discussed earlier, on the whole portmanteau forms rather than suffixed inflection, and are thus phonologically less If a grammatical category/property-based distinction is likely to lose inflection through the reduction of endings lost in noun phrases, the evidence from the Germanic common in adjectives, nouns and verbs in the Germanic languages indicates that the distinction is also eventually languages. Furthermore, the high degree of suppletion lost in the personal pronouns. Personal pronouns cannot in the personal pronouns means that given phonological indefinitely uphold a grammatical category/property- reduction, forms which have a suppletive distinction will

13 Diachronic change in the real world/grammatical basis of categories is also possible, for example in grammatical to natural gender where selection of the gender pronoun becomes increasingly governed by the gender of the real-world referent rather than the grammatical gender of the antecedent. This development – attested to varying extents in English and other Germanic languages – contradicts the hypothesis of unidirectionality proposed in grammaticalization theory (e.g. Hopper & Traugott 1993, chapter 5).

（ 9 ） ― ―10 福岡大学研究部論集 A ９（７） 2009

tend to remain formally distinct longer than those with tor formatives is very efficient when a larger number of less suppletive distinctions. A further factor is that fre- meanings is involved. A few hundred morphemes may quent forms are acquired by children as individual forms be combined into an enormous number of messages. Yet before the acquisition of general patterns. the memory load – and the learning load – is relative to The distinction between grammatical and real-world- the flexibility obtained for the system. This kind of sys- based categories/properties made above is important tem, therefore, is efficient in its use of large open classes here. Although it cannot be expected, indeed the evi- of forms’ [Pike’s italics]. For an ‘ideal’ (i.e. suppletive dence from the Germanic languages shows, that a gram- portmanteau) matrix, Pike states ‘The efficiency here is matical category/property-based distinction lost in noun very great in terms of compactness of signal, since two (or phrases will be maintained indefinitely in the personal more) categories are carried by the single – often short pronouns – their delay or maintenance of the form can be – formative.’ He adds, ‘although the single-celled forma- explained by the reasons discussed – the maintenance of tive is highly efficient in these terms, it must be harder to some real-world-based category/property distinctions in learn and remember, however, specifically because of the the personal pronouns, such as person or natural gender, complexity which gives it that efficiency. This memory even when absent from noun phrases, may be to facilitate load seems to put some kind of a limit on the number of clearer reference.14 Note the occurrence of other real- such formatives – e.g. affixes and particles – which any world-based categories/properties in pronouns to facili- one language can maintain. For this reason it is only in tate reference: in English and other Germanic languages small closed systems (such as a pronominal set)17 that personal/non-personal and animate/inanimate often one is likely to find extensive use of single-celled forma- come under the heading of natural gender, and proximity tives.’ is a category in this–that, as it is in Japanese kore–sore– However, although this does indeed seem to describe are. Similar categories outside the Germanic languages much of the inflection in large and small word classes in and Japanese include inclusive/exclusive or visible/invis- for example English – the personal pronouns are general- ible.15 ly more suppletive than many larger word classes – there To summarize, then, grammatical distinctions lost in seems to be no purely numerical reason why a large word noun phrases are eventually lost in personal pronouns, class should necessarily be morphologically regular. In but some real-world distinctions may be maintained to the largest ‘word class’ of all, the lexicon, thousands facilitate reference. Maintenance of clear reference – of lexical items with little or no formal connection are both grammatical and real world – will be discussed in learnt with no apparent difficulty, and it is quite possible the following section in ambiguity as a factor in change – indeed in the world as a whole quite common – for lan- in the personal pronouns. guage-users to learn two or more languages, increasing the size of the vocabulary even further. In addition, not Therapeutic change only do speakers memorize individual forms, but their lexical entry may also specify accent or tone, (in Ger- Several authors have commented on the importance of manic languages for instance) gender (of nouns) and (in the size of the word class in determining the type of German for example) a non-predictable plural marker, as morphology or degree of regularity–irregularity, and the well as for many language-users a written form – in Eng- example of the personal pronouns has often been given lish and Japanese frequently a non-predictable spelling as an illustration of the type of morphological system or kanji. Therefore, the suggestion that a suppletive word found in a relatively small word class.16 For example, class must be small because of a numerical difficulty in Pike (1965: 205f.) states ‘The “simple” matrix, with vec- remembering a large number of forms seems not to hold

14 T/V, on the other hand, is socially deictic. 15 Maintenance of some real-world-based category/property distinctions in the personal pronouns to facilitate clearer reference also accounts for the absence of animacy or personal/non-personal distinction in the 1st and 2nd person pronouns. 16 Whether Japanese has a personal pronoun word class, and the size of any such word class, will be discussed further in Howe (forthcoming a). 17 We will look at closed-classness in a later paper, but can state here that new pronouns can be and indeed have been added to this supposedly ‘closed class’, even in English, and certainly pronouns can be lost from this ‘closed’ class. We should rather state that the class of pronouns in many languages – indeed the class of function words – is comparatively stable, i.e. new forms can be added, but much less often than is the case for lexical or content words (see Howe 1996: 100–104 and forthcoming b).

（ 10 ） Pronoun Morphology（Howe） ― ―11

– it seems quite within the capacity of a language-user to communicative situation. learn a large number of formally unconnected items. Ambiguity also depends on the frequency of use of homonymic forms – i.e. if they are comparatively infre- Ambiguity quent, then ambiguity is unlikely to be a problem, even if they can occur in identical contexts. The importance However, where the size of the word class may be rel- of ambiguity and therapeutic change in the personal evant is in the toleration of homonymy: it is likely that pronouns argued here can be connected with the fact homonymy will be tolerated much less in a relatively that the personal pronouns are function words, and have small word class such as the personal pronouns where both high frequency of use and a condition of referential forms with very similar functions and reference occur non-ambiguity – as stated above, the personal pronouns frequently in similar or identical contexts. Forms such as are a small set of forms with very similar functions and English hair/hare, time/thyme, vein/vain/vane – or Japa- reference occurring frequently in similar or identical nese hana (flower)/hana (nose) – are unlikely to occur in contexts. Ambiguity in the personal pronouns concerns the same contexts where ambiguity could be a problem. speakers and hearers in communicative situations, i.e. the Similarly, grammatically as well as semantically forms speaker–hearer interface and speaker–hearer interaction, such as threw (V)/through (P) – down (P, V)/down (N) and as such is change motivated in speech and by com- – blue (A)/blew (V) – an (det)/Anne (N) – and indeed I/ municative need. eye/aye/(the letter)I – you/yew/ewe/(the letter)U – we/ wee – him/hymn – mine/mine – ours/hours – are unlikely Changes to suppletion to occur in the same context and result in a homophonic clash, i.e. in ambiguity. In the small word class of the Morphologically, relatively few of the therapeutic highly frequent personal pronouns, however, homonymy changes in the pronouns are by the addition of regular is much more likely to be a hindrance to comprehension. inflection. One reason for this is that often the personal In Howe (1996), I argue for the importance of thera- pronouns have, as discussed earlier, comparatively little peutic change in the personal pronouns in the Germanic regular, consistent inflectional pattern and consequently languages. Many developments in the pronouns can be often there is very little inflectional pattern in the person- explained as a remaking of category/property distinctions al pronouns to follow. Furthermore, there is often little which were still valid but which had become ambiguous. or no appropriate noun phrase pattern to follow either Example causes of ambiguity in the personal pronouns as some real-world distinctions (such as person) made are phonological merger, merger through T/V usage (for in personal pronouns are absent in noun phrases and, in example 2nd p. plural used as 2nd p. singular), or loss of ambiguity in nominative singular forms, the nominative former disambiguating verb morphology. singular in noun phrase inflection may be unmarked/ As pointed out by Gilliéron,18 homonymy in itself markerless for case, number and/or gender. However – as does not lead to ambiguity, rather it is where homonymic shown by for example English plural you-s(e) – where a forms occur in the same contexts that a homonymic clash pattern does exist, changes in the pronouns may follow or ambiguity can result. Note also that unlike ‘homony- this pattern. my’, i.e. formal sameness, which is a fixed concept, am- Rather than by regular inflection, a number of the biguity is a variable. ‘Ambiguity’ thus allows differences changes in the personal pronouns in the Germanic lan- from language to language, such as the significance of guages show a type of therapeutic change akin to the word order, verb morphology or context for example, to lexical replacement discussed by Gilliéron on the basis be accounted for. In addition, a variable term ‘ambigu- of the Atlas linguistique de la France – i.e. by a complete ity’ also allows for the possibility of category/property change of form – though in this case a pronominal form. hierarchy (nominative over oblique for instance) and for Where in Gilliéron’s examples therapeutic change by accented–unaccented use, and further can include on an suppletion is by lexical replacement, or in English go– ambiguity scale addition of quantifiers such as us two, went by verbal replacement, in the personal pronouns de- you all, or they both, which may also be defined as clari- velopments that show therapeutic change by suppletion fications aimed at facilitating the task of the hearer in the show pronominal replacement – i.e. distinct forms are

18 For a summary of Gilliéron see for example Bynon (1977: 186–190).

（ 11 ） ― ―12 福岡大学研究部論集 A ９（７） 2009

taken from the pronouns themselves.19 Not all therapeu- As discussed in the introduction, the personal pro- tic change in the pronouns involves the use of redundant nouns are generally (co)referring terms, both grammati- forms for repair; however, use of redundant forms is seen cally and semantically to the external world – in their in oblique pronouns as subj. forms (as in Swedish 3rd p. core meaning ‘I’ = the speaker, ‘we’ = a group to which plural dom), in dual pronouns as plural forms (as possi- ‘I’ belong, ‘he’ = the male person etc. – and therefore it bly in Icelandic við and þið), in generalization of distinct is perhaps not surprising that also formally the personal variant forms (as possibly in English she), and in the bor- pronouns show similarities both with regular morphol- rowing of foreign or dialect forms (as in English they). ogy and with lexical or content words. The occurrence of For further discussion and examples, see Howe (1996: suppletive portmanteau forms in the personal pronouns, chapter 2). as well as the fact that they must be learnt individually, The systematicness of suppletive morphology in that although not the rule in the grammar, is not exceptional it marks distinctions between forms sharing the same at all in language as a whole. category/property or categories/properties – i.e. forms are distinct from those with which they share a category/ References property or categories/properties – as opposed to marking correspondences as with patterning morphology – has Bolinger, D. L. (1948) ‘On Defining the Morpheme’, Word already been discussed above, and here it is argued how 4/1, 18-23. a number of developments because of ambiguity – i.e. a Brown, R. & Gilman, A. (1960) ‘The pronouns of power and lack of adequate distinction – can and do result in sup- solidarity’, in T. A. Sebeok (ed.), Style in Language, pletion – i.e. the taking of forms to make a distinction. MIT, pp. 253–276. This thus demonstrates one further way that suppletive Bunt, Jonathan (2003) Japanese Grammar and Verbs, Oxford: morphology in the personal pronouns can arise. The sys- Oxford University Press. tematicness of such developments and the frequent lack Bynon, T. (1977) Historical Linguistics, Cambridge: Cam- of morphological parallel show how developments that bridge University Press. result in suppletive distinction cannot simply be regarded Carstairs-McCarthy, A. (1992) Current Morphology, London as ‘irregular’. & New York: Routledge. Crystal, D. (1991) A Dictionary of Linguistics and Phonetics, Conclusions 3rd ed., Oxford: Blackwell. Crystal, David (2003) The Cambridge Encyclopedia of Lan- As well as correspondences between form and meaning, guage, 2nd ed., Cambridge: Cambridge University the personal pronouns show in their degree of suppletive Press. morphology a correspondence between form and func- Forchheimer, P. (1953) The Category of Person in Language, tion. Not only can there be a connection between form Berlin: De Gruyter. and meaning in the sense of consistent category/property Fromkin, V. & Rodman, R. (1993) An Introduction to Lan- correspondences, but there can also be a connection be- guage, 5th ed., International edition, Fort Worth: Har- tween form and function in the greater degree of supple- court Brace Jovanovich. tive morphology as a result of the factors discussed in Gimson, A. C./Ramsaran, S. (1989) An Introduction to the this paper. Pronunciation of English, 4th ed., London: Edward The personal pronouns are only ‘irregular’ in the sense Arnold. that they do not conform to the usual pattern of inflection Greenberg, J. H. (1966) Language Universals with Special in the language. Cf. Werner (1991: 396) who states ‘In- Reference to Feature Hierarchies, The Hague: Mou- stead of postulating a basic uniformity in language which ton. can be “disturbed” …, we should look for a consistent Greenberg, J. H. (ed.) (1978) Universals of Human Language, and comprehensive theory of language change that ex- III, Wordstructure, Stanford: Stanford University plains this non-uniform behaviour.’ Press.

19 Lexical replacement or avoidance may occur not only because of ambiguity, but also because of taboo – see e.g. Lehmann (1992: 90, 260, 263f.) for examples. As touched on in an earlier footnote, to some extent taboo is also a factor in pronoun usage in the Germanic languages – see Howe (1996: §0.1.3) – and represents the extreme end of a scale of pragmatic factors influencing pronoun usage. It is also a factor in pronominal usage and avoidance in Japanese – for a discussion of taboo in Japanese pronouns, see Suzuki (1978: 123).

（ 12 ） Pronoun Morphology（Howe） ― ―13

Haas, W. (1966) ‘Linguistic Relevance’, in C. E. Bazell et al. Matthews, P. H. (1974) Morphology. An Introduction to the (eds), In Memory of J. R. Firth, London: Longmans, Theory of Word-Structure, Cambridge: Cambridge Green & Co., pp. 116-147. University Press. Head, B. F. (1978) ‘Respect Degrees in Pronominal Refer- Mühlhäusler, P. & Harré, R. (1990) Pronouns and People: ence’, in Greenberg 1978: 151-212. The Linguistic Construction of Social and Personal Hinds, John (1986) Japanese, Routledge: London and New Identity, Oxford: Blackwell. York (= Croom Helm Descriptive Grammars Series) Nübling, D. (1992), Klitika im Deutschen. Schriftsprache, Hockett, C. F. (1987) Refurbishing our Foundations, Amster- Umgangssprache, alemannische Dialekte, Tübingen: dam/Philadelphia: Benjamins (= Amsterdam Studies Gunter Narr Verlag (= ScriptOralia 42). in the Theory and History of Linguistic Science. Se- Pike, K. L. (1965) ‘Non-linear Order and Anti-redundancy ries IV, Current Issues in Linguistic Theory 56). in German Morphological Matrices’, Zeitschrift für Hopper, P. J. & Traugott, E. C. (1993) Grammaticalization, Mundartforschung 32, 193-221. Cambridge: Cambridge University Press. Quirk, Randolph; Greenbaum, Sidney; Leech, Geoffrey; Howe, Stephen (1996) The Personal Pronouns in the Ger- Svartvik, Jan (1985) A Comprehensive Grammar of manic Languages: A Study of Personal Pronoun the English Language, London and New York: Long- Morphology and Change in the Germanic Languages man. from the First Records to the Present Day, Berlin and Shibatani, Masayoshi (1990) The Languages of Japan, Cam- New York: Walter de Gruyter (= Studia Linguistica bridge: Cambridge University Press. Germanica 43). Suzuki, Takao (1978) Words in Context. A Japanese Perspec- Howe, Stephen (2009) ‘Irregularity in Pronouns’, paper pre- tive on Language and Culture, English translation by sented at the conference on Irregularity in Morphol- Akira Miura, Kodansha International: Tokyo. (Japa- ogy (and Beyond), Bremen, Germany, October 2009. nese original = Kotoba to Bunka, Tokyo: Iwanami Howe, Stephen (forthcoming a) ‘Personal Pronouns in Eng- Shoten, 1973). lish and Japanese: A Preliminary Comparison’. Svartvik, J. et al. (1982) Survey of Spoken English. Report on Howe, Stephen (forthcoming b) ‘Pronoun Categories, Rein- Research 1975-81, Lund (= Lund Studies in English terpretation and Sources of New Pronoun Forms in 63). English and Japanese’. Tsujimura, Natsuko (2007) An Introduction to Japanese Lin- Howe, Stephen (forthcoming c) ‘Reference and Ellipsis in guistics, 2nd edition, Malden, MA/Oxford: Black- English and Japanese’. well. Howe, Stephen (forthcoming d) ‘Pronouns and Politeness in Visser, W. (1988) ‘In pear klitisearringsferskynsels yn it English and Japanese’. Frysk’, in S. Dyk & G. de Haan (eds), Wurdfoarried Ingram, D. (1978) ‘Typology and Universals of Personal Pro- en wurdgrammatika: in bondel leksikale stúdzjes, nouns’, in Greenberg 1978: 213-248. Ljouwert/Leeuwarden (= Fryske Akademy 692), pp. Lehmann, W. P. (1992) Historical Linguistics, 3rd ed., Lon- 175-222. don & New York: Routledge. Werner, O. (1991) ‘The Incorporation of Old Norse Pronouns Makino, Seiichi and Tsutsui, Michio (1986/1989) A Diction- into Middle English. Suppletion by Loan’, in P. S. ary of Basic Japanese Grammar, Tokyo: The Japan Ureland & G. Broderick (eds), Language Contact in Times. the British Isles. Proceedings of the Eighth Interna- Mańczak, W. (1980) ‘Frequenz und Sprachwandel’, in H. tional Symposium on Language Contact in Europe, Lüdtke (ed.), Kommunikations-theoretische Grund- Douglas, Isle of Man, 1988, Tübingen: Niemeyer, pp. lagen des Sprachwandels, Berlin/New York: De 369-401. Gruyter, pp. 37-79.

（ 13 ）